Smads are a small family of eukaryotic transcription regulators that play key roles in the transforming growth factor-beta signaling cascade. Smad6 and Smad7, the inhibitory or I-Smads, inhibit signaling downstream of TGF-beta type I receptors, thereby acting as negative regulators of signaling mediated by TGF-beta superfamily of ligands. Smad6 is known to specifically inhibit BMP type I receptor mediated signaling, while Smad7 is a more general inhibitor, able to block signaling mediated by a set of related TGF-beta type I receptors, including type I receptors for BMP and TGF-beta/Activin. In this study we have sought to understand the structural basis for this functional divergence of I-Smads. We have created homology-based models for the MH1 and MH2 domains of the two I-Smads and have carried out detailed molecular dynamics (MD) simulations of these proteins in explicit solvent to investigate the flexibility of the domains. The molecular models show that the I-Smads have lost many of the secondary structural elements found in the R-Smads, giving rise to longer loops in the tertiary structure of Smad6 and Smad7. Detailed analysis of the structural models and the MD trajectories clearly reveal that compared to Smad6, Smad7 has a more flexible overall folding, marked by the presence of highly flexible amino acid residues in functionally important regions of the protein. Interestingly, three of these residues-Phe411, Lys401, and Cys406, map to L3 loop of Smad7 MH2 domain, which is a critical structural determinant in Smad-type I receptor interactions. The increased structural flexibility of Smad7, arising out of longer, more flexible loops in its MH2 domain, might enable Smad7 to interact with a set of related yet structurally diverse type I receptors. Taken together with experimental evidence published in recent literature that hint at structural factors underlying the generic nature of inhibition by Smad7, our results strongly suggest that structural flexibility could be a prime contributor to the functional differences between Smad6 and Smad7. Additionally, we have been able to use the Smad7 structural model to successfully rationalize the results of in vitro site-specific mutagenesis experiments in published literature. This also provides biological validation for our model. Apart from this, analysis of the MH1 molcular model of Smad6 delineates a basic patch on the surface of the domain that might take part in nonspecific DNA binding by Smad6. This finding is consistent with earlier experimental data and is relevant since the characteristic beta-hairpin DNA binding element of R-Smads is completely absent in the I-Smads. Finally, the molecular models described here can serve to guide future biochemical and genetic studies on I-Smads.
(c) 2008 Wiley-Liss, Inc.